Gamma-Ray-Anchored 1D Mechanical Earth Modeling in Sparsely Logged Horizontal Wells: Application to Hydraulic-Fracture Design

Abstract

Reliable geomechanical modeling for hydraulic-fracture design in horizontal laterals is constrained by limited data availability. The full petrophysical log suites necessary for accurate modeling increase rig time by 2–4 days and are often impractical to obtain in highly deviated wells. As a result, gamma ray (GR) measurements become the sole continuous data source in most completed laterals. Existing data-scarce approaches such as regional averaging and generic GR correlations provide insufficient accuracy for reliable geomechanical modeling. While machine learning (ML) techniques have potential, they are constrained by the scarcity of representative training data sets in data-limited environments. Recent measurement-while-drilling (MWD) GR workflows capture relative stiffness variations. However, they depend on pilot-well arrays and 3D seismic data sets to establish calibration relationships, infrastructure that is absent in many mature and emerging unconventional plays. We present a deterministic single-point calibration workflow that overcomes these limitations using only one calibration depth. Our single-point anchor approach provides absolute-magnitude constraint absent from regional averaging and generic GR correlations while eliminating the need for pilot wells or seismic-derived calibration models. Seven well-specific petrophysical parameters measured at the calibration depth are combined with the continuous GR log to generate pseudologs through deterministic transformations. These pseudologs are then used to construct an internally consistent 1D mechanical Earth model (MEM). The proposed workflow is benchmarked against regional averaging and generic GR correlations under identical hydraulic-fracture simulation conditions. It is validated using eight wells across three geologically distinct basins representing siliciclastic, limestone-shale, and chalk-marl lithologies. Single-point calibration achieved a mean maximum surface-treating-pressure error of 2.46% across six Spraberry-Wolfcamp wells. Errors were 1.52% in an independent blind test of the Buda Formation, with all logs withheld, and 0.3–5.1% in a Niobrara validation case. Every stress barrier within the validation data set was correctly resolved, and no calibrated well exhibited predicted out-of-zone (OOZ) fracture growth. Two cases marginally exceeded the 5% error threshold, and both had net-pay exceeding 200 ft. This suggests a preliminary performance boundary in highly heterogeneous interbedded systems, for which net-pay thickness may serve as a practical proxy pending further validation. In contrast, both uncalibrated baseline approaches failed. Regional averaging predicted caprock breach in every well, whereas generic GR correlations produced four distinct failure modes, including complete vertical misplacement that surface-treating pressure alone could not detect. Pseudolog generation requires no proprietary software, facilitating implementation in data-limited unconventional reservoirs.